[Vlavlbausers] Error in AIPS task FILLM for polarimetry data

[Data Analysts]

To: All VLA users
Subject: Error in AIPS task FILLM for polarimetry data

Dear colleague,

We have recently discovered an error in the AIPS task FILLM whereby
cross-polarization data (LR and RL) from the VLA may be scaled and
weighted incorrectly, in a manner that cannot be corrected by calibration.
The specific instances of when this incorrect scaling and weighting occurs
are described below.  The 31DEC08 (TST) version of FILLM was corrected
on March 5, 2008, and is available for download via the midnight job.
A patch is also available for AIPS version 31DEC07 (NEW).  Please note
that 31DEC06 (OLD) has not been patched, and should no longer be used
to load VLA data into AIPS.

More details:

FILLM has been scaling cross-hand data incorrectly since the Modcomps
were turned off on June 27, 2007, for data saved by the on-line system in
raw correlation coefficients but loaded scaled by the nominal sensitivity
(the default option).  This problem occurs on those baselines for which
the antenna order in the on-line system is reversed from that required by
AIPS (e.g., where the on-line system provides a visibility for baseline
11-6 rather than 6-11).  Once the data have been loaded into AIPS it
is not possible to tell which baselines may have undergone this switch.
The problem also affects data obtained in the Modcomp-era that are loaded
as correlation coefficients by FILLM, where FILLM has had to "undo" the
on-line nominal sensitivity scaling.  Thus all solar data taken with
the Modcomps, which were always converted to correlation coefficients
by FILLM, have had their cross-hand data incorrectly scaled.

FILLM has also been calculating the weights for the same cross-hand data
incorrectly, although this has a much less serious effect on the data.

The fix for this problem is to:

1) re-load data using a corrected version of FILLM in 31DEC07 or 31DEC08

2) copy all calibration and flag tables from any previous reduction
to the new dataset, and check that the data were not overly flagged
previously due to poor cross-hand results

3) re-do any polarization processing

Parallel-hand (RR and LL) data reductions are unaffected, so most
total-intensity experiments will be fine.

Please contact daip at nrao.edu if you have any further questions about
the quality of your data.

Claire Chandler
Deputy Assistant Director for Science, NM Operations